Night-light detection of squid and light-luring fishing fleets
VIIRS Day-Night Band imagery locates squid and light-luring fishing fleets at night by detecting their high-intensity lamps from low Earth orbit, exposing fleets that routinely disable AIS and evade SAR during calm conditions.
Sensors
- VIIRS DNB (Suomi NPP): Day-Night Band panchromatic channel, 500 m spatial resolution at nadir, single daily overpass near 01:30 local time, detectable radiance floor around 2×10⁻¹¹ W/cm²/sr. Operational since 2012, providing a decade-plus archive for trend analysis.
- VIIRS DNB (NOAA-20): Identical DNB specification to Suomi NPP, launched 2017, offset orbital plane giving a second nightly overpass roughly 50 minutes apart. Together the two satellites allow same-night confirmation of fleet positions and reduce single-pass cloud-loss risk.
- Luojia-1: Chinese experimental night-light satellite, approximately 130 m ground resolution in its panchromatic night-light band, far finer than VIIRS DNB. Revisit is irregular (roughly monthly at any given location) and archive depth is limited, but it can resolve individual vessel separations that VIIRS merges into a single blob.
- VIIRS Nightfire (EOGDATA product): A processed derivative of VIIRS shortwave-infrared and DNB data produced by the Colorado School of Mines. Applies Planck-curve fitting to estimate source temperature and radiance, distinguishing vessel lights from gas flares and other thermal emitters. Published nightly at approximately 500 m.
Why squid boats are bright enough to see from 830 kilometres up
Jumbo flying squid (Dosidicus gigas), Argentine shortfin squid (Illex argentinus) and Japanese flying squid (Todarodes pacificus) are strongly phototropic. Fishing vessels exploit this by deploying arrays of metal-halide or LED lamps, often totalling 50 to 300 kilowatts per vessel, to concentrate squid near the surface at night. The resulting radiance is not subtle. A single well-lit squid jigger can outshine a small town when viewed from orbit.
VIIRS DNB was designed to detect moonlit clouds and city lights, with a noise-equivalent radiance of roughly 2×10⁻¹¹ W/cm²/sr. A 200 kW lamp array at sea surface produces radiances several orders of magnitude above that floor, making detection straightforward on cloud-free nights. The practical limit is not sensitivity but spatial resolution: at 500 m per pixel, a cluster of vessels within roughly 500 m of each other merges into a single detection. Luojia-1's 130 m pixels can partially resolve that problem, though its infrequent revisit makes it a supplement rather than a backbone.
Separating lamp-light from moonlight, city glow and gas flares
Raw DNB radiance on any given night is a mixture of sources: vessel lights, moonlight reflected from the sea surface, atmospheric scatter from coastal cities, and thermal emitters such as offshore gas flares. Separating fishing vessels from this background requires several correction steps that are well-established in the published literature.
Moonlight correction uses the lunar irradiance model and the known lunar phase and elevation angle to predict the expected sea-surface reflectance contribution pixel by pixel. Residuals above a radiance threshold, typically set empirically for each ocean basin, are candidates for artificial light sources. A second filter cross-references VIIRS shortwave-infrared bands (I4 and I5, centred near 3.7 µm and 11.45 µm) against the DNB signal: gas flares produce strong thermal emission in those bands, while cold lamp arrays do not. The VIIRS Nightfire algorithm from the Colorado School of Mines formalises this separation using Planck-curve fitting across multiple bands.
Cloud masking is the most stubborn operational limit. Optically thick cloud completely obscures vessel lights. In persistently cloudy fishing grounds such as the Yellow Sea in summer or the Patagonian shelf in austral winter, cloud-free DNB observations may be available on fewer than 40 percent of nights. Multi-night compositing recovers fleet presence over a week or month but cannot establish a vessel's exact position on a specific night.
What the fleet distribution actually tells you
A single VIIRS overpass locates lit vessels to within roughly 500 m and timestamps their position to within a few minutes of the 01:30 local overpass. That is enough to establish which exclusive economic zone (EEZ) a vessel is operating in, whether it is within a marine protected area, and whether the fleet density is consistent with licensed effort levels reported by the flag state.
Repeated nightly observations build a fleet-movement picture that AIS alone cannot provide. Vessels that disable their transponders to avoid scrutiny still need their lights on to catch squid. The NOAA-20 and Suomi NPP overpasses, separated by roughly 50 minutes, also allow a crude speed estimate for vessels that move detectably between passes, though 500 m resolution and the short time window limit this to flagging fast movers rather than precise velocity measurement.
Fleet counts derived from DNB have been validated against AIS-reporting vessels in well-monitored fisheries, providing a ratio of detected-to-reporting vessels that can be applied to estimate total effort including the dark fraction. Published work on the North Pacific and South Atlantic squid fisheries suggests that the AIS-dark fraction can exceed 50 percent in some seasons and areas, though the precise figure varies by fishery and year.
The archive matters as much as the live feed
Suomi NPP VIIRS data runs back to October 2011. NOAA-20 adds a second track from early 2018. That combined archive supports multi-year trend analysis: which fleets expanded into new areas, which EEZs saw effort spikes coinciding with stock assessments, and whether bilateral fishing agreements produced measurable changes in fleet presence. For fisheries enforcement agencies, a decade of nightly positions is often more useful than a real-time alert, because it establishes the baseline against which anomalies are judged.
The EOGDATA Nightfire product from the Colorado School of Mines makes much of this archive publicly accessible, though extracting fleet-specific intelligence from it still requires filtering, georeferencing to EEZ boundaries, and cross-referencing with vessel registries and AIS records. Satellize runs this pipeline operationally, and the same analytical approach that underpins the Tonga crop-estimation programme applies here: open sensor data processed through documented methods, with outputs calibrated to the client's specific geographic and regulatory context.
Honest limits and what they mean for operational use
Cloud is the central constraint and there is no satellite fix for it. In persistently overcast regions, DNB-based monitoring must be treated as a probabilistic record of fleet presence rather than a nightly census. Combining DNB with SAR (which sees through cloud but cannot detect unlit vessels in calm seas) partially compensates, but each method has its own failure modes and neither is a complete substitute for the other.
Vessel separation at 500 m resolution means that a dense cluster of 20 jiggers anchored within a square kilometre may register as two or three detections rather than twenty. Fleet counts in high-density fishing grounds should therefore be treated as lower bounds. Luojia-1 imagery, when available, can refine the count, but its irregular revisit means it cannot be relied upon for time-sensitive operations.
Finally, DNB detects light, not vessel identity. Matching a DNB detection to a specific vessel requires cross-referencing with AIS, vessel monitoring system (VMS) data if the flag state shares it, or SAR detections from the same night. Without that cross-reference, DNB tells you a vessel was present and approximately where, not who it was or what flag it flew.
Typical figures
| Spatial resolution (VIIRS DNB) | ~500 m at nadir; degrades to ~750 m at swath edge |
| Spatial resolution (Luojia-1) | ~130 m (panchromatic night-light band) |
| Revisit (Suomi NPP + NOAA-20 combined) | Two overpasses per night at most locations, separated by ~50 minutes, near 01:30 local time |
| Spectral band (DNB) | Panchromatic, 500–900 nm; VIIRS Nightfire also uses I4 (~3.7 µm) and I5 (~11.45 µm) for source-type discrimination |
| Minimum detectable radiance (DNB) | ~2×10⁻¹¹ W/cm²/sr (noise-equivalent radiance) |
| Global swath width (VIIRS) | 3,040 km, achieving near-daily global coverage |
| Archive depth | Suomi NPP from October 2011; NOAA-20 from early 2018 |
| Data latency (EOGDATA Nightfire product) | Typically available within ~24 hours of overpass |
| Cloud limitation | Optically thick cloud completely blocks detection; cloud-free fraction can fall below 40% in persistently overcast regions |
| Delivery formats | GeoTIFF radiance grids, GeoJSON vessel-detection point files, CSV fleet-count time series |
Analytics Satellize can run
| Nightly fleet position map | Radiance thresholding on moonlight-corrected DNB, with VIIRS Nightfire gas-flare mask applied | GeoJSON point layer of detected lit vessels, timestamped to overpass, with radiance value per detection |
| EEZ incursion alert | Spatial join of DNB detections against authoritative EEZ boundary polygons (e.g. Maritime Boundaries Geodatabase); flagging of detections inside closed areas or MPAs | Automated alert report listing detection coordinates, EEZ attribution and distance to nearest boundary |
| Dark-fleet fraction estimate | Ratio of total DNB detections to AIS-reporting vessels in the same area and time window, using published validation ratios from comparable fisheries | Monthly summary table of estimated total effort versus licensed effort, with confidence range |
| Multi-year fleet-effort trend | Time-series aggregation of nightly detection counts from the VIIRS archive (2011 to present), normalised for cloud-cover frequency per cell | Annual effort-density raster and trend chart per fishing ground, suitable for stock-assessment input |
| Fleet-movement trajectory (dual-pass) | Matching of detections between Suomi NPP and NOAA-20 overpasses (~50-minute separation) to flag vessels with significant positional displacement | Flagged detection pairs with displacement vector and implied minimum speed, delivered as GIS layer |
| Cloud-gap-filled weekly composite | Maximum-radiance compositing over rolling seven-night window to recover fleet presence through intermittent cloud cover | Weekly GeoTIFF composite with per-pixel observation-count layer indicating data confidence |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.